Abstract Study question Can non-invasive PGT (niPGT), using a protocol that avoids manipulation of the embryo before spent culture medium is collected, accurately determine embryo chromosomal status? Summary answer Accuracy using a truly non-invasive PGT method was low, suggesting this strategy is not appropriate for clinical use and should be restricted to research studies. What is known already PGT requires embryo biopsy to obtain small numbers of cells for genetic testing. The procedure requires skilled personnel and specialist equipment, significantly adding to the cost of PGT while also creating training and logistical bottlenecks that reduce patient access to PGT. Additionally, there have been concerns that biopsy could damage embryos. These considerations have stimulated interest in potential non-invasive PGT strategies. Thus far, niPGT methods have mostly focused on cell-free embryonic DNA found in spent culture media, but studies reporting highest accuracy have tended to involve embryo manipulations likely to promote DNA release and therefore cannot be considered truly non-invasive. Study design, size, duration Samples of spent culture medium (SCM) associated with 128 embryos were collected. Oocytes had been denuded of cumulus cells, fertilised using ICSI and the resulting embryos thoroughly washed on day-4 before transfer to a fresh drop of medium. SCM samples were collected on day-5 or day-6 of culture. Embryos had not previously been cryopreserved and underwent minimal manipulation prior to SCM collection. Results of SCM analysis were compared to those obtained following conventional ‘invasive’ PGT. Participants/materials, setting, methods Media samples were subjected to whole genome amplification using the PG-Seq Rapid Non-Invasive PGT kit. Concentrations of sequencing libraries were measured, and next-generation sequencing (NGS) was undertaken. The resulting NGS data was analysed with PG-Find Software to predict chromosomal status. In parallel, trophectoderm biopsies from the corresponding embryos were analysed using a well-established and highly validated PGT method, based upon amplification of thousands of sites across the genome, followed by NGS. Main results and the role of chance Despite the implementation of rigorous measures to prevent contamination of SCM samples, extraneous female DNA, likely of cumulus cell origin, was frequently observed. Depending on the clinic, 16.7% to 38.1% of media samples associated with male embryos gave a discordant (female) result. This highlights the difficulty of avoiding inadvertent sampling of maternal DNA. There were no instances of male DNA contamination affecting SCM samples from female embryos. We assessed whether clinic of origin, ploidy status of the embryo, and day of SCM collection influenced the quality of the sequencing libraries (ANOVA). Higher concentrations of libraries and superior sequencing quality scores were observed in SCM collected on day-6 compared to day-5 (p = 0.011 and p = 0.002, respectively). Chromosomal assessments obtained from SCM were compared to those from standard, trophectoderm-based PGT in a blinded fashion. 50% of SCM samples gave an identical predicted karyotype to their corresponding trophectoderm biopsy (concordant copy number for all chromosomes). When applying a simple aneuploid/euploid classification, 80% of embryos classified chromosomally abnormal using the validated PGT method were also aneuploid according to analysis of the associated SCM sample, but without necessarily having identical karyotypes, while 62% of embryos classified euploid by PGT received the same designation after SCM analysis. Limitations, reasons for caution SCM samples were designated ‘contaminated’ when embryos of one sex (determined using invasive PGT) were incorrectly classified as the opposite sex using niPGT. However, our study was unable to determine when SCM samples were contaminated with DNA of the same sex as the embryo. Consequently, contamination rates could be higher. Wider implications of the findings Most niPGT methods are not truly non-invasive, as they include embryo manipulations that increase likelihood of cell death and DNA release. Our results show that obtaining reliable niPGT results is difficult without deviating significantly from optimal embryological protocols. Currently, the accuracy of niPGT seems insufficient to justify routine clinical use. Trial registration number Not applicable